Reconciling Temperature Trends Generated from Raw Versus Homogenised Maximum Temperature Datasets for Darwin, Australia, 1895 to 2014
Author: Jennifer Marohasy
Received: 2 September 2016
Accepted: 4 November 2016
Key words: data quality control, control charts, temperature time series, homogenisation, Stevenson screen, ACORN-SAT, Southern Oscillation Index
Abstract: Homogenisation of raw temperature data, measured from mercury thermometers in weather stations, is routinely undertaken before calculating global warming trends. This is ostensibly to correct for non-climatic factors. However, the use of ill-defined algorithms and comparative sites, can make the method difficult to independently verify or replicate. Darwin’s temperature record is important because this is the only location in central northern Australia where temperatures have been measured since 1895 from within a Stevenson screen. The homogenised series for Darwin have been contested, in particular adjustments that change a cooling trend in the raw data into statistically significant warming. Through the application of Minitab statistical software, in particular I-MR-S (individual, moving range and standard deviation) control charts, raw and homogenised maximum temperature time series for Darwin are reanalysed. Temperature trends in the raw maxima for Darwin are shown to be broadly synchronous with other locations in north eastern and north western Australia, including Richmond, the only location with a long continuous record from 1895 to the present. We show that the adjustment in the official Australian Climate Observation Reference Network – Surface Air Temperatures (ACORN-SAT), where the maximum temperature is stepped-down by -1.12°C from 1937 back to 1910, to be inappropriate. In particular, the abrupt drop in observed temperature maxima at Darwin, ostensibly because of shading from trees, occurred after 1937. We hypothesise that this cooling was not due to shading, but rather a cyclone in March 1937 clearing away vegetation that had previously screened the post office from the sea breeze. We create a new minimally-homogenised maximum temperature series for Darwin, correcting only for the move to the airport in February 1941. This temperature series for Darwin shows no overall cooling or warming temperature trend for the period 1895 to 2014, consistent with the trend for Richmond.
Please cite this article as: Marohasy, J., Reconciling Temperature Trends Generated from Raw Versus Homogenized Maximum Temperature Datasets for Darwin, Australia, 1895 to 2014, New Climate (2016), https://doi.org/10.22221/nc.2016.002
For a prepublication copy of this manuscript please contact NevilleWhite@climatelab.com.au